Scratch Build mini

[chimera15]

Damn, $95 for a set of blanks.  That's dumb.  I mean, Unicomp sells them for $20 (which you might even be able to argue is high itself).  They can't just throw a few caps in bag and send them your way for around $20?

Yeah I know, but the problem is because it's relatively custom, and has a lot of odd shapes.  If anyone knows anyone selling cherry caps cheaper that would fit this layout let me know.

[itlnstln]

You can grab a set of Filco caps (blank or w/legends) for about $30 + shipping from Elitekeyboards.  It won't help you for the custom stuff, but at least it'll give you a bunch to work with.

[chimera15]

You can grab a set of Filco caps (blank or w/legends) for about $30 + shipping from Elitekeyboards.  It won't help you for the custom stuff, but at least it'll give you a bunch to work with.

I have a bunch to work with from the scorpius..but  I'd rather not make this board contingent on the availability of m10's alone, or Filcos for that matter.   Not to mention I'd rather have blank caps than black on white. So Majestouch sells blank black cherry caps for a tenkeyless for $30?  Hmm.

[Findecanor]

It's a real shame I can't use this plate.  The plate is perfect except for the right shift, which is in precisely the wrong place for me to put a shrunken right shift and another arrow key.  If it wasn't for that one bad section I could cut down the plate and use it as is.

Show Image

I still think that you could cut that board apart and weld parts of it together. It would be a little bit fiddly work to get the parts aligned perfectly straight, but I think it would still be much, much easier than to drill and dremel a plate from scratch.
For the right shift and cursor keys, could you use /-*-8-9-6-3-, on the numpad? For the left Ctrl/Alt/Fn/Fn, could you use Right Ctrl-Left-Down-Right ?

[chimera15]

I'm kind of wanting to learn how to do a cnc drawing/production piece right now, since I've never done it and it'll be useful to me in the future.  I also think I should be able to punch the holes for the pcb with a cnc as well. Hopefully it won't be too costly, and it'll be exactly correct.

[chimera15]

Look at this.  Melissa wrote me back.  Only 5 sets, the price drops by 70%??  Wow. 5 sets would only be $150...wth??  Only 1.5x the cost of 1 set for 5!??  Funky pricing...

It also appears the bottom keys are the major cost.

"I don't yet have the list of the enter keys, I will send that tomorrow when I get back in the office, but volume pricing for the blank set would be as follows:

1 set             $95 / set

5 sets           $30 / set

10 sets          $23 / set

15 sets          $20 / set

20 sets          $19 / set

 

If you were to only order the bottom row of keys (we actually call this Row4) - I am assuming 9-single sized keys and 4 larger sized keys including the space bar, volume pricing for that would be:

1 set             $85 / set

5 sets           $20 / set

10 sets          $12 / set

15 sets          $10 / set

20 sets          $8 / set

 

I know one set seems like alot, but when you spread that setup time over a larger quantity of keys, the unit price goes down.

Again, we are talking blank keys - no printing whatsoever, and this does not include freight/shipping.

 

Let me know if you have any questions - I will get that list of enter keys over to you tomorrow.

Thanks!"

If I buy enough to actually build kits/boards to sell, and assuming that the pcb/plate/case don't get too pricey that means it drops the price to HHKB prices by that fact alone.... This might actually be feasible.

The thing I'm worried about right now is the case, but I need to design the plate and pcb first and get those taken care of.

[chimera15]

I finished an illustrator file with exact measurements.  All I need is to know what enter key size they have.  Was really tough to get all the measurements correct.  This is a jpg version of the vector file:



I realize I need to come up with a better name for these drawings if I'm going to show them to people.  I can't just keep calling it a scratch built mini, and I need to put information on the drawing for what it is.  



Any suggestions? I was thinking about "GeekHack Hacking Keyboard"?  So GHH, since DMW and a lot of other people have contributed to this board so far. Would that be acceptable, or should I think of something that doesn't associate with this forum...? lol  Let's make a keyboard company like the KBT guys. lol

edit: I realize looking at that I made a 3mm mistake on the left spacebar stabilizer hole.  Well it's still a wip.

[chimera15]

Fixed the stabilizer hole.

[lowpoly]

Nice, you could convert that to iges and import into the emachineshop software. I'll stop mentioning emachineshop now. :-)

Here's a CNC punch:

http://www.youtube.com/watch?v=IAHU0a03OPM

So your plate might be possible without special tools.

[chimera15]

Nice, you could convert that to iges and import into the emachineshop software. I'll stop mentioning emachineshop now. :-)

Here's a CNC punch:

http://www.youtube.com/watch?v=IAHU0a03OPM

So your plate might be possible without special tools.

Really interesting. That does look like what I need.  My next step is to design the pcb wiring.  Then I'll probably try to put them all together and design a case in a 3d program, so I can figure out where the holes will go for screws as well.

[chimera15]

Surprisingling or rather not based on what DMW was telling me, SP doesn't sell the little stabilizer plastic pieces for Cherry caps.  They expect you to use more than 1 switch to stabilize a large cap, which seems unreasonable on a production board.  Anyone know where I can get those little plastic parts and maybe even the wires?  I can cannibalize the scorpius for the first board, but if I actually manage to produce more than a handful of these it'll become necessary to find a place that sells them.  If I can't find a place I may have to redesign the board for extra switches.

[chimera15]

I realized I put too many keys in the drawing. lol  It's got an extra set of keys. roflol

[chimera15]

That's better:

SP is telling me they don't have the right size enter key really, so might have to chance the layout some...

[dmw]

Yeah, it really sucks to have to be limited to only those keys that they already have tooling for.  I'm typing full-time now on my HH1, and I'd really like to have the double-height enter keys that I had originally intended before finding out it would cost me >$1500.

As for the stabilizer bars - the double keyswitch method works well in practice.  I have no problem with mine.  And there's another benefit: your PCB could have mounts for three adjacent switches, and the end-user could decide on the arrangement of the caps.  For example:  1 long + 1 short, 1 short + 1 long, or 3 short.

[dmw]

Really interesting. That does look like what I need.  My next step is to design the pcb wiring.  Then I'll probably try to put them all together and design a case in a 3d program, so I can figure out where the holes will go for screws as well.

I did my case in Google Sketchup, which I found really easy to work with.  The only difficult part was getting an output file that Shapeways would accept.  I had to do a lot of conversions.

Before going to Shapeways, I considered having some plexiglas custom-cut, and just sandwiching the keyboard between two pieces.

[chimera15]

Worked all day on these.  This is my wiring solution, such as it is.  I noticed DMW used a much more logical method, but not 100% sure it would work in my case since I'm using cherry's + alps and he just used alps, so there isn't quite as much room.  It's pretty cluttered and high tolerance.  Not sure how much luck I'd have at making some kind of pattern like this by hand.  I think I may have to research photo-etching to do this.  

This was the best I could come up with for the traces.





The major concern I have is that some traces are going to be running an extremely long distance with this configuration.  I'm wondering if the resistance will be too high at the end of the run to recognize the activation of the switch...

The second of course is that the traces may come into contact with the solder points as I have to have them running close to the holes/contacts and under the switches.

This was like doing a puzzle on steroids. Or maybe a maze. lol, and this was simple as far as matrix goes.  I don't know how they ever figure it out for a normal chip.

Last step in illustrator is to do a cap layout which I need for Signature Plastics.  So far they've been pretty good.  An order of 5 sets of caps is pretty realistic, but they need to know the exact measurements of the odd shaped keys.

I realize I need to offset one of the alps pin holes still.   Alps will be mounted upside down.  I also need to do a plate alternative for them.  For now I'm going to concentrate on making a cherry board though.

[dmw]

Nice start!  Here are a few of my thoughts:

• you could greatly simplify this if you route all your columns on one side of the board, and all your rows on the other.  Even with mounts for both Alps and Cherry switches, you should have plenty of room for the traces.
• you'll have much better results if you use PCB drawing software rather than Illustrator for this kind of work.  Besides, you'll need to if you want to send these off to be made.
• Maybe it might be better to ditch the teensy++ and just go with the straight components.  Trying to layer a teensy over the keyboard really complicates things.  Look at the AT90USB chip on the teensy, it's ... well it's teensy.  You could put it down in the blank area where the spacebar goes.

I've only used ExpressPCBs software (which is extremely easy to use), though I've dabbled with a few others.  If you want to give it a try, I'll be happy to lend you a hand.  There's also kicad.  I'm going to do the next version of my board using that, so I'm not tied to only one board vendor.

[ricercar]

This is great work. Helpful and clearly documented. I'm taking notes for my own point-to-point construction.

[chimera15]

Nice start!  Here are a few of my thoughts:

• you could greatly simplify this if you route all your columns on one side of the board, and all your rows on the other.  Even with mounts for both Alps and Cherry switches, you should have plenty of room for the traces.
• you'll have much better results if you use PCB drawing software rather than Illustrator for this kind of work.  Besides, you'll need to if you want to send these off to be made.
• Maybe it might be better to ditch the teensy++ and just go with the straight components.  Trying to layer a teensy over the keyboard really complicates things.  Look at the AT90USB chip on the teensy, it's ... well it's teensy.  You could put it down in the blank area where the spacebar goes.

I've only used ExpressPCBs software (which is extremely easy to use), though I've dabbled with a few others.  If you want to give it a try, I'll be happy to lend you a hand.  There's also kicad.  I'm going to do the next version of my board using that, so I'm not tied to only one board vendor.

I don't think I have the budget to send them out to be done for the price you quoted.  Unless they have a deal like Signature plastics where they actually sell more for less...

I think I'm going to have to do it myself.  I don't think its going to be impossible.

I didn't see any other way to do the layout without extending the size of the pcb... If I put the chip directly on the board then it becomes something that I definitely can't do on my own, and I'm not sure that putting the chip near the spacebar is possible without making the traces very small, which they are on the chip already.   It would surely increase the price.  I'm hoping there's an easy way to do photo etching.  I think I read about a simple method once to do it.  Maybe it was on Mythbusters even, when they did that thumb print episode...

[chimera15]

Yeah from my old model making days:

http://www.starshipmodeler.com/tech/fh_pe.htm

http://www.fullnet.com/~tomg/gooteepc.htm

Ok, need a laser printer sounds like.

[chimera15]

Laserjets are laser printers right?  I think this one would work?

http://cgi.ebay.com/HP-4000N-LASERJET-PRINTER-/140422615209?cmd=ViewItem&pt=COMP_Printers&hash=item20b1d710a9

Cheaper one:

http://cgi.ebay.com/HP-2100-LASERJET-PRINTER-/140422614651?cmd=ViewItem&pt=COMP_Printers&hash=item20b1d70e7b


Yup looks like :

"- Print the pattern, using the darkest laser printer settings (On my LaserJet 4, I use: Dithering: NONE, Intensity: DARKEST, "Raster Graphics", "Print Truetype as Graphics", and RET (Resolution Enhancement Technology): OFF). If you have a "Manual Feed" tray, such that the paper doesn't have to bend as many times, or as much, USE IT. [But I still sometimes have to "help" the paper to keep moving, as it exits the printer. (Maybe my rollers are dusty?)] "

[lowpoly]

IIRC, these guys have been recommended here before:

http://batchpcb.com

[dmw]

Wow.  BatchPCB looks great.  My board, if done by them, would cost ~$165!  That's about half of what I paid ExpressPCB.  They said in their FAQ that costs could go as low as 5 cents per square inch on large runs.  That would bring my board down to $3.30!  Though I don't think I'll ever order enough volume to get that kind of discount, it gives me a better idea of the possible price range.

[dmw]

I don't think I have the budget to send them out to be done for the price you quoted.  Unless they have a deal like Signature plastics where they actually sell more for less...

I think I'm going to have to do it myself.  I don't think its going to be impossible.

Definitely not impossible.  But you are still better off using some kind of PCB software, so if you ever do decide to do a production run, you'll be all set.  Plus, the software helps you get your design right. The right tool for the job and all that...

I didn't see any other way to do the layout without extending the size of the pcb...

Did you try putting the column traces on the same side as the teensy board?  I think you'll find that the row traces are much simpler this way.  If you want to try redoing your layout in kicad or one of the other free PCB CAD tools, I'll be happy to help with your layout.

If I put the chip directly on the board then it becomes something that I definitely can't do on my own, and I'm not sure that putting the chip near the spacebar is possible without making the traces very small, which they are on the chip already.

Those tiny chips can be soldered by hand, but you're right, it is much more difficult.  I'm not sure why putting the chip near the spacebar would affect the size of the traces.  The traces would have to start small anyway since the leads on the chip are tiny.  Traces that small can be a problem however when etching your own board.  You especially want to avoid sharp corners.  So maybe it is better to stick with the Teensy, at least for the first version.

It would surely increase the price.

Board prices seem to be measured in square inches, not in number of components.

[chimera15]

It's a real problem.  Your post made me consider why what you said wouldn't work.  It brought up another problem with a two sided design...  The switches have to be mounted on the other side from the chip the way I have it currently designed.  As well as what you suggested.  Right now the chip is blocking those switch positions, which I hadn't considered.   I think I may have to extend the pcb or figure out what you're talking about.  One way or another what you're suggesting will cause me to have to put jumps for leads into the pcb, from the backside to the front and such, and I didn't want to do that if I didn't have to.

I may be able to put a little module on top of the keyboard or something.  If I keep the teensy++ board that means though that the connection will now be parallel to the board unless I desolder the port and mount it somehow in the case, which kind of sucks.

The other way might just to make solder attachments that I could just wire up and bring the chip back over to the other side and mount in the case....

[dmw]

Ah, I didn't realize that you were attempting to design a single-sided board.  I assumed both of the designs you posted were to be etched, but the side with the teensy, those traces are just wires, right?  It's making more sense to me now.

Making a double-sided board by hand would be a chore, I believe.  You would have to have both sides lined up really well, so when you drill your holes you hit the right point on both sides.  I've never done this mind you.  I'm just speculating.

If you want to stick to a single sided board, you could shorten those crazy row traces by adding short jumper wires to your design.   Something like this:



The jumper wires are in red.  I don't know if you had diodes in your design, but I've left them out of this simple mock-up.

[chimera15]

My board is double sided, just I don't want to use through the board traces, or jumpers wires.  That's the problem.  Every one of those jumps will take time to solder in, that's 16x5 =80 jumps, which will be time consuming and ridiculous to do. I'd rather just be able to do the process and dunk it in the acid, then wire the chip and switches, and the 21 through the board jumps already on the board.

[dmw]

Ah, sorry, I get it now.  I didn't realize that your overriding constraint was minimizing soldering by minimizing through-holes and jumpers.  With that in mind, your board designs make a lot more sense to me now.

Are you planning on hand-building a lot of these?

[dmw]

Just a thought, but maybe, since the teensy board is only 0.7 inches wide, you could mount it sideways at a board edge using 90 degree angle pin headers.  One edge would be a regular short 90 degree header, and the other edge would be a long one.  I found some long ones online that might fit the bill.



Might work?

[chimera15]

Just a thought, but maybe, since the teensy board is only 0.7 inches wide, you could mount it sideways at a board edge using 90 degree angle pin headers.  One edge would be a regular short 90 degree header, and the other edge would be a long one.  I found some long ones online that might fit the bill.

Show Image

Might work?

Yeah, the big problem though is that if I mount it any other way than it is right now in the design, the connector will be difficult to access.  I think I'm going to have to create through the board holes and use wires so I can mount it in the case.  The really big problem of doing it that way is that it doesn't even have a hole to put a screw in it to keep it tied down.  What a mess.

[dmw]

You could use this:



Or alternately, don't expose a plug at all. Just add a place in the case to secure the cable before it exits.  Wrap it around a screw post or something.

Another option would be to mount a mini-B plug directly to the main board, and run wires from the teensy to it.  Just 4 more wires and a plug to solder.  Not too bad.

[dmw]

You know, with those three methods of running a cable, there's no need to have the teensy rigidly attached to the main board.  You could use solid wire ribbon cable and attach the teensy directly to the edge of the main board, then fold it over so it lays flush with the board.

[chimera15]

You know, with those three methods of running a cable, there's no need to have the teensy rigidly attached to the main board.  You could use solid wire ribbon cable and attach the teensy directly to the edge of the main board, then fold it over so it lays flush with the board.

Yup, that's what I was planning, one way or the other.

[chimera15]

So here's the new idea.  The chip is suspended below the circuit-board and tied to it by wires.  

[chimera15]

That step may actually be unnecessary though looking at the back side.   I could probably just simplify it by running the wires directly to the holes instead of using secondaries now.  I may be able to get away with a single side circuit board

[didjamatic]

Seriously awesome thread.

[chimera15]

Alright, the final idea, with the wire layout, and the chip directly hooked up to the trace side by wires.  Then the other side will be completely blank with the switches on that side and all the wiring and traces on the other.  



I'll have to come up with something in the case that will allow the wiring to be routed cleanly.  Also it occurred to me there should be a hole in the case that will allow the pushing of the programming button on the teensy so it can be programmed without breaking open the case.

[dmw]

Some questions and suggestions:

You'll want to go through and eliminate all the sharp bends in your traces.  They tend to get eaten during etching, leaving you with very thin corners.

The 4 traces running between the alps mounts on the top row look awfully tight. Do you have any actual measurements for the trace widths and the distance between them?  What about the pads?  How much copper are you leaving around the holes?

Just curious - how do you plan on drilling the 400+ holes?

[chimera15]

Some questions and suggestions:

You'll want to go through and eliminate all the sharp bends in your traces.  They tend to get eaten during etching, leaving you with very thin corners.

The 4 traces running between the alps mounts on the top row look awfully tight. Do you have any actual measurements for the trace widths and the distance between them?  What about the pads?  How much copper are you leaving around the holes?

Just curious - how do you plan on drilling the 400+ holes?

Yeah I know, it's screwed, and further screwed now by the fact that I measured the standard key which turned out to be 18mm, with a hole of 14mm for each switch it means there's 2mm on each side, which means my measurement of 4mm between each hole is wrong, which meant my whole board is off, and it must be a 5mm gap between each switch.  I measured that gap like 20 times.  I had to reset everything to 5mm between each switch hole.

Here's the cap layout:



I need to get some hard  data on the hole measurements now to sort out all the other problems now, but it's super tough since it's about radius's and angle from a centerpoint.  I should have started with this first.

[chimera15]

Yeah my bottom row looks offcenter right now as well...it may be if I shift it over that that right shift becomes normal size? hmm...

This is really odd...

It becomes really difficult when I start measuring in inches because it becomes about fractions of inches, and you have to start adding fractionally.

For instance, the center of the first key is at x=.47 inches, y=1.74 inches  The second key on the right is at x=1.22 inches  So what's that...75 right? but tons of math I have to punch into a calculator.  Well that was a handy check yeah, so 5 mm sounds right.  I need to do that check on the second and third row, thanks.  I wish you had it in mm. lol

How wide is a key Ripster?

I don't get it, this is really confusing me.  Everything looks right in the layout, but the keys look off because the middles don't line up right.

[chimera15]

This should make a little more sense I hope:


Alright, thanks to Ripster's measurements I should have it a lot closer if not perfect.

[chimera15]

Does anyone know above what measurement you should add stabilizers for a cherry key by chance?

[chimera15]

I was able to copy the cherry schematic into illustrator, then use it as an underlay to get the position of the drill points exact.  As well I was able to drill through my m10 board and experiment where an alps switch would sit perfectly. So I should be able to tell from that close enough where the alps holes will fit.  Surprisingly it looks like I had more room than I originally thought.  The big question will be if there will be enough room for good round soldering pads.   It may be that I just do the wiring that goes straight through the hole, then you would have to solder the switch to that point, which might be harder to solder.  I'll try to get some good pads there.

[chimera15]

Redid all the wiring.



I think that's it for illustrator.  I also added solder points around the drill holes, but they're pretty thin relatively. One solution to this might be to extend the trace around this point into directions that the wiring isn't conflicting.  This should provide a good anchor for the switch, and not get in the way of the traces.   I hope I can use the same cnc I use to do the plate to drill them.  The next step I think is to go 3d, then I have to break down and actually spend some money I guess. lol

The largest question with this layout of course won't be known till it's made, that if the resistance is too great in the 5th row and cause problems.  The longest trace to left control is equivalent of being at least several feet long.

[dmw]

Looking better!

Some considerations:

I notice that there are no diodes in your layout.  This would mean that you'd have to turn on ghost-key blocking in the firmware.  With a strict grid layout as you've done, you're likely going to run into problems with your modifier keys. For keyboards without diodes, some thought usually goes into how to arrange the matrix so that modifier keys don't cause ghost-key blocking.  Dave Dribin's site has a good description of the problem, and the solution with diodes.

If the trace length does turn out to be a problem, you could shorten them somewhat by running half the rows over the top of the column traces and the other half (or 2/5ths as the case may be) around the bottom of the column traces.  This can be accommodated by moving the through-holes for the columns up from the bottom slightly - maybe even between the keys.  This would also help you avoid the problem I mentioned before, with too many traces crunched between the holes for the Alps key mounts in the top row.

[chimera15]

Looking better!

Some considerations:

I notice that there are no diodes in your layout.  This would mean that you'd have to turn on ghost-key blocking in the firmware.  With a strict grid layout as you've done, you're likely going to run into problems with your modifier keys. For keyboards without diodes, some thought usually goes into how to arrange the matrix so that modifier keys don't cause ghost-key blocking.  Dave Dribin's site has a good description of the problem, and the solution with diodes.

If the trace length does turn out to be a problem, you could shorten them somewhat by running half the rows over the top of the column traces and the other half (or 2/5ths as the case may be) around the bottom of the column traces.  This can be accommodated by moving the through-holes for the columns up from the bottom slightly - maybe even between the keys.  This would also help you avoid the problem I mentioned before, with too many traces crunched between the holes for the Alps key mounts in the top row.

Hmm, I suppose I should test that on my test board...  I didn't notice it being like 2 key rollover or anything.  That may be something that can be done later, but I'm not sure there's enough room there as it stands to drill more holes.  If I can actually get a plate/board made maybe you can show me how to turn the anti-ghosting feature on on the teensy.

[dmw]

Ghost-key blocking is turned off on your test board.  So you won't see 2 key rollover, but you will see ghost characters.

It looks to me like there is room for diodes between the rows.  Especially if you reroute the last two row traces around the bottom instead of the top.

[dmw]

maybe you can show me how to turn the anti-ghosting feature on on the teensy.

Just edit Matrix.kspec and change:

Code: [Select]

BlockGhostKeys:no
to

Code: [Select]

BlockGhostKeys:yes

[chimera15]

Melissa from Signature Plastics and I finally got the layout worked out for the caps, so I'm going to order 5 sets.  She also redid the drawings in illustrator for the caps.  Pretty amazing:

[chimera15]

Now I just need a plate which is my next top priority for this board.  The blank pcb still hasn't come from the person I ordered it from....I may need to find a new source.